Shifters raises 10.2M$: AI-driven ground robots for high-risk missions

Shifters raises $10.2M: AI agent-driven ground robots for high-risk missions

🔎 "The robot should enter before the human" — why this funding round changes the game

On June 3, 2026, Shifters, an Israeli deep-tech startup, announced a $10.2 million seed round led by Ace Capital Partners. A sum that seems modest compared to the hundreds of millions injected into humanoid robots, but which targets a specific niche: autonomous quadruped robots for environments that humans should no longer enter first.

Founder Ofer Ballin's formula sums it all up: "The robot should enter before the human." Behind this statement, an entire paradigm is taking shape. That of Physical AI applied to the field, with onboard edge computing and multi-agent coordination. Not a robotic arm fixed to an assembly line. A system that moves, decides, adapts.

The timing is not coincidental. Agentic AI is stepping off screens to incarnate in physical machines. Models like OpenAI's GPT-5.5 (agentic score of 98.2 on reference benchmarks) open the door to planning and reasoning capabilities that did not exist just 18 months ago. Shifters takes these advances and injects them into ground robotic platforms designed for real-world chaos.

The key points

• Shifters raises $10.2M in seed funding (Ace Capital Partners), bringing its total funding to $15M.
• The startup develops autonomous quadruped robots powered by Physical AI and edge computing for defense and hazardous environments.
• The goal: to coordinate teams of multi-agent robots capable of reconnaissance and maneuvering in complex terrain without continuous human intervention.
• The funding is intended to strengthen the autonomy technology, improve the design, and prepare for manufacturing.

Recommended tools

Who is Shifters — the genesis of an Israeli deep-tech startup

Shifters was founded by Ofer Ballin and Assaf Chaprak. The company is based in Israel and positions itself as a player in "AI-native defense robotics", according to the terms of The Jerusalem Post.

The startup does not build a generic robot. It develops autonomous robotic teams organized around a multi-agent architecture. Each quadruped robot is equipped with sensors and autonomy software designed for maneuvering in difficult terrain.

The approach stands out from aerial drones, which have become widely democratized. The ground is a fundamentally different problem: three-dimensional obstacles, unstable surfaces, lack of direct line of sight. Shifters focuses exclusively on this specific complexity.

The "Robots Go First" positioning is not a marketing slogan. It is an architectural choice: every design decision starts from the principle that the robot will be deployed in conditions where a human should not go.

Physical AI technology — what sets Shifters apart from classic robots

Physical AI: when the LLM meets the real world

Physical AI refers to the integration of advanced AI models into systems that physically interact with their environment. This is not improved teleoperation. It is autonomous reasoning applied to movement.

At Shifters, this approach translates into onboard edge computing. The robot does not rely on a cloud connection to make decisions. It processes its sensor data locally, runs AI models directly on board, and coordinates its actions with the other robots in the team.

This is where recent advances in agentic AI become crucial. A model like GPT-5.5, with its score of 98.2 on agentic benchmarks, is capable of breaking down a complex mission into sub-tasks, anticipating obstacles, and readjusting its plan in real time. Transposed into a ground robot, this changes the very nature of autonomy.

Edge computing and zero latency

The choice of edge computing is not a technical luxury. It is an operational necessity. In military or hazardous industrial environments, communications can be jammed, intermittent, or unavailable.

According to TechTime, Shifters robots integrate edge computing capabilities that allow them to operate autonomously even when the connection with the command post is lost. The robot does not become useless. It continues its mission with the parameters it received.

Multi-robot coordination

The most fascinating aspect of the Shifters platform is multi-agent coordination. It is not about deploying an isolated robot, but teams that communicate with each other, divide up coverage areas, and collectively adapt their behavior.

Military AI reports that the funding will specifically be used to advance this agentic dimension for ground robots: coordinated reconnaissance, management of dangerous missions, and multi-robot orchestration. This is exactly the type of architecture described in our analysis on agentic AI for robotics.

The robotic defense market — why quadrupeds now

The aftermath of recent conflicts

The conflicts of the 2020s have accelerated the adoption of robotic systems by armies worldwide. Drones have had their demonstration moment. But the ground remains the most dangerous and least automated domain.

Human losses during reconnaissance missions in urban areas or rough terrain have pushed the military to demand autonomous ground solutions. Not teleoperated ones — these require constant bandwidth and a dedicated operator. True autonomous systems.

This is the niche Shifters is attacking. And they are not the only ones, but their AI-native approach sets them apart.

Quadrupeds vs humanoids: the right tool for the right job

The debate between quadrupeds and humanoids is raging. Humanoids like those from Figure 02 capture media attention. But for military missions in chaotic terrain, the quadruped offers structural advantages: stability on four support points, the ability to get back up after a fall, and a reduced footprint.

A quadruped can cross rubble, climb steep slopes, and move in confined spaces where a humanoid would struggle to maintain its balance. For the "go in first, scan, report" mission, it is often the best compromise.

AI systems like Helix de Figure AI, on the other hand, target the general public and structured environments. Shifters targets the opposite: the unpredictable, the dangerous, the unstructured.

Funding as a market signal

The fact that Ace Capital Partners is leading this raise is significant. Ace Capital is known for its investments in Israeli defense technologies. According to Yahoo Finance, this $10.2M raise brings Shifters' total funding to $15M, with global investors involved.

For a seed-stage robotics startup, $15M is enough to move from prototype to pre-series. It is also a clear signal: the autonomous robotic defense market is maturing.

Breakdown of the raise — where the money goes

Announced allocation

According to Defense Daily, the funding will be allocated to three priority areas:

1. Strengthening autonomy technology: improving onboard AI models, optimizing edge computing, developing real-time agentic reasoning capabilities.
2. Design improvement: mechanical robustness, sensor integration, field deployment ergonomics.
3. Manufacturing preparation: setting up the production line to move from demonstration units to the first operational batches.

CityBiz confirms that the goal is to "scale" the AI-powered terrestrial robotics platform for defense, security, and hazardous industrial operations.

What this allocation tells us about maturity

The allocation is revealing. A startup that would devote the bulk of its seed to pure R&D would still be at the conceptual stage. Shifters is already allocating resources to manufacturing. This means the prototype is functional and scaling is the next bottleneck.

The fact that autonomy remains the top expense item, however, indicates that the core technology is not yet set in stone. Agentic AI is evolving fast — too fast to lock into a given architecture.

Skynet Watch — what should concern us

Lethal autonomy: the thorny question

Shifters denies being in the realm of Lethal Autonomous Weapons (LAWS). Official communication talks about "reconnaissance" and "dangerous missions." But the line between a reconnaissance robot and an armed robot is technologically thin.

When a system is capable of navigating autonomously, identifying threats, and coordinating a response with other agents, the question is no longer "can it kill?" but "who decides that it kills?". Agentic AI adds a layer of complexity: with models like Claude Opus 4.7 Adaptive (score 94.3) or Gemini 3 Pro Deep Think (95.4), the reasoning capacity is real. The traceability of decisions, however, remains an open problem.

Multi-agent coordination as a risk multiplier

An autonomous robot is already a complex machine to audit. A team of robots coordinating their actions in a decentralized way is a system whose emergent behavior is fundamentally unpredictable.

Imagine five quadruped robots on a reconnaissance mission. The onboard agentic model decides that one of them must bypass a building on the left while another covers the entrance. This decision is made in milliseconds, based on sensor data and a neural model whose weights are not trivially interpretable.

Who is responsible if this maneuver leads to a misidentification? The operator who gave the initial mission? The model developer? The commander who deployed the team?

The semantic shift

Observe the evolution of language in the sector. We have gone from "teleoperation" to "supervision" to "supervised autonomy" to "autonomy with human-in-the-loop intervention." Each semantic shift corresponds to an effective withdrawal of the human from the decision-making process.

Shifters speaks of "supervised autonomous robots" according to Embodied Global. The word "supervised" is reassuring. But in practice, human supervision of multi-agent systems operating in hostile terrain with communication latencies often boils down to an emergency stop button — not effective control.

The access asymmetry

15 million dollars is the budget of a seed startup. The R&D budgets of national armies are counted in the billions. The technology Shifters develops today will tomorrow be in the hands of state actors with infinitely superior means to deploy it at scale.

The question is not "is Shifters dangerous?" The question is: "When this technology is mature, who will have it and under what control?"

AI Models and Robotics — What Agentic Scores Actually Mean

From Screen to Field

The agentic scores we track (GPT-5.5 at 98.2, Gemini 3 Pro Deep Think at 95.4, Claude Opus 4.7 Adaptive at 94.3) measure a model's ability to plan, execute, and adjust complex chains of action. In robotics, these capabilities translate concretely.

A high score means that the model can: break down a reconnaissance mission into sequential steps, anticipate points of failure, reallocate resources (in this case, other robots) if an agent is taken out of action, and generate structured reports in real time.

Self-Hosted Models: A Sovereignty Issue

The value of self-hosted models like Kimi K2.6 Moonshot AI (88.1) or GLM-5 Reasoning from Z.AI (82) takes on its full meaning in a military context. Deploying a model via OpenAI or Google's API means that mission data passes through third-party servers. Unacceptable for defense.

Shifters has not publicly detailed which model its robots embed. But the edge computing architecture strongly suggests the use of models optimized for local inference, potentially compressed variants of open-weight models or proprietary models trained on specific field data.

The Role of AI in the Decision Loop

The typical architecture of a system like Shifters' looks like this: sensors (LiDAR, cameras, IMU) feed a perception model that builds a representation of the environment. This representation is processed by the agentic model which generates an action plan. The plan is executed by the motor controller. The loop repeats at 10-50 Hz depending on the complexity of the terrain.

This is AI in every link of the chain. And that is exactly what makes auditability so difficult.

Shifters facing competition — who wins the ground robot race

The competitive landscape

The autonomous ground robot market for defense is structured around several players with different approaches.

Shifters differentiates itself through its strictly AI-native and multi-agent approach right from the design phase. Many competitors added autonomy in successive layers to platforms that were initially teleoperated. Shifters views autonomy as the foundation, not as an add-on.

The Israeli advantage in defense tech

Israel holds a unique position in the defense tech ecosystem. The country combines an experienced military sector (operational needs are concrete and constant), a mature VC ecosystem for deep-tech, and close relationships with international defense markets.

According to The Jerusalem Post, Shifters fits into this tradition of Israeli defense innovation, with investors like Aurelius who specifically understand this market.

The risk of too narrow a niche

The danger for Shifters is locking itself into a military niche that limits its addressable market. Defense is a demanding but capricious client: sales cycles are long, certifications are heavy, and political changes can call everything into question.

CityBiz mentions that Shifters also targets "hazardous industrial operations" — mines, nuclear plants, chemical sites. This is smart. But the startup's storytelling remains centered on defense, which could slow down diversification.

What this fundraise means for the robotics industry

The signal for VCs

A $10.2M seed round for an AI-native ground robotics startup is a clear signal: defense-focused VCs are betting on Physical AI. Not on drones (that's already been done). Not on consumer humanoids (too early). On autonomous ground robots for specific missions.

The timing with the maturation of agentic models is no coincidence. Investors see the convergence between software AI (agentic LLMs) and physical AI (robots) and are betting on startups that combine both from the design stage.

The acceleration of the shift from R&D to operations

The fact that Shifters is already allocating funds to manufacturing indicates that the industry is entering a concrete phase. The years 2023-2025 were those of lab demonstrations. 2026-2027 will be those of the first real operational deployments.

For armies, this changes everything. Moving from a demonstration prototype to a robot that can actually be sent on a mission requires mechanical, logistical, and software reliability that does not exist in a lab.

The ripple effect on the ecosystem

Each fundraise of this type creates a ripple effect. Engineers working at Shifters acquire Physical AI expertise that will spread throughout the ecosystem. Sensor suppliers, embedded computing providers, and mechanical component manufacturers benefit from the demand. Other startups find it easier to raise because the market is validated.

❌ Common mistakes

Mistake 1: Confusing autonomy and enhanced teleoperation

Many commentators equate Shifters' robots to enhanced versions of teleoperators. This is fundamentally wrong. The agentic architecture with edge computing means that the robot makes local decisions without waiting for instructions. Teleoperation is a fallback mode, not the default mode.

The distinction is crucial for assessing the risks and actual capabilities of the system.

Mistake 2: Judging a ground robot by the same criteria as a humanoid

Comparing Shifters to Figure AI or Agility Robotics makes no sense. They do not solve the same problem. Shifters' quadruped is optimized for stability and robustness in unstructured terrain. The humanoid is optimized for interacting with an environment designed for humans. Choosing one or the other depends on the mission, not on the "quality" of the robot.

Mistake 3: Ignoring the question of auditability

Focusing solely on technical performance (speed, battery life, obstacle-climbing capability) without addressing the traceability of the AI's decisions means missing the main subject of controversy. A robot that works perfectly but whose decisions cannot be explained poses a major legal and ethical risk.

Mistake 4: Assuming that $15M is enough for production

15 million dollars is significant for a seed round. It is insufficient for military-scale production. This raise is a ticket to entry for a Series A, not a sufficient starting balance. Certification costs, real-world testing costs, and production deployment costs far exceed this budget.

❓ Frequently Asked Questions

What types of missions exactly does Shifters target?

Reconnaissance in complex terrain, assessment of dangerous areas (UXO, unstable structures), perimeter surveillance, and missions where sending a human in first presents an unacceptable risk. The startup does not detail specific scenarios for operational reasons.

What exact AI model does Shifters use?

Not publicly specified. The edge computing architecture suggests models optimized for local inference, potentially compressed variants. The use of self-hosted models like Kimi K2.6 or GLM-5 would be consistent with data sovereignty requirements in a defense context.

Does Shifters build armed robots?

No public information indicates this. Sources exclusively mention reconnaissance and "dangerous missions." However, the quadruped platform is compatible with diverse payloads, and the technical distinction between a reconnaissance robot and an armed robot is thin.

Who are the target customers?

Israeli defense forces first and foremost, then international defense markets via partnerships. Shifters also mentions dangerous industrial sectors (mining, chemical, nuclear) as secondary markets.

How does Shifters' funding compare to other robotics startups?

$15M total is on the higher end for a pure robotics seed round (excluding consumer humanoids which raise astronomical amounts). This is consistent with a startup that already has a working prototype and is preparing for manufacturing.

✅ Conclusion

Shifters embodies the concrete convergence between agentic AI and physical robotics: $10.2M for quadruped robots that think, coordinate, and act on their own in hostile environments. The technology is real, the market is valid, and the ethical questions are urgent. The robot goes in before the human — but who decides what it does once inside?